Chapter 15: Forecasting

In this chapter:
1. Forecasting chicken consumption using OLS (UE 15.1, Equation 6.8, p. 501)
2. Forecasting chicken consumption using a generalized least squares (GLS) model estimated
with the AR(1) method (UE 15.2.2)
3. Forecasting chicken consumption using a generalized least squares (GLS) model estimated
with the Cochrane-Orcutt method (UE 15.2.2)
4. Forecasting confidence intervals (UE 15.2.3)
5. Forecasting with simultaneous equation systems (UE 15.2.4)
6. Forecasting with ARIMA models (UE 15.3)
7. Exercises

Forecasting chicken consumption using OLS (UE 15.1, Equation 6.8, p. 501):

The chicken demand model developed in Chapter 6 was estimated using data from 1951-1994. In
order to forecast a variable beyond 1994, the workfile range and sample must be expanded.
Follow these steps to forecast chicken consumption for 1995 - 1997 using ordinary least squares:

Step 1. Open the EViews workfile named Chick6.wf1.
Step 2. To expand the workfile range, select Procs/Change Workfile Range from the workfile
menu bar, change the End date from 1994 to 1997, and click OK.
Step 3. To expand the sample, select Sample on the workfile menu bar, change the second number in
the window from 1994 to 1997, and click OK. After Steps 2 & 3 are completed, the top portion
of the EViews workfile should
look like the graphic on the
right (note that the second
date for Range: & Sample:
have changed from 1994 to
1997).
Step 4. To open Y, PC, PB, & YD in a group window, hold down the Ctrl button, click on Y, PC, PB,
& YD, select Show from the workfile toolbar, and click OK.
Step 5. To enter the data for Y, PC, PB, & YD from the text & footnote number 4 in UE, p. 502 (see
table below), click edit+/- on the group menu bar, scroll to the bottom of the spreadsheet and
replace the NA's with the appropriate numbers for the period 1995 - 1997. Press Enter after each
entry and click edit+/- on the group menu bar a second time to save your changes. Scroll to the
bottom of the group spreadsheet and make sure that it looks like the table below.

Y PC PB YD
1995 80.30000 6.500000 61.80000 200.6200
1996 81.90000 6.700000 58.70000 208.5000
1997 83.70000 7.700000 63.10000 216.3100

Step 6. Select Objects/New Object/Equation on the workfile menu bar, enter Y C PC PB YD in the
Equation Specification: window, change the Sample: to 1951 - 1994, and click OK.
1

Step 7. Select Name on the equation window menu bar, enter EQ01 in the Name to identify object:
window, and click OK.
Step 8. Select Forecast on the equation menu bar, enter YFOLS in the Forecast name: window, set
the Sample range to forecast: to 1951 1997, and click OK.
Step 9. Open YFOLS & Y in a group window by holding down the Ctrl button, clicking on YFOLS
& Y, selecting Show from the workfile toolbar, and clicking OK. Scroll to the bottom of the
group spreadsheet and make sure that it looks like the table below. Note that the predicted values
for YFOLS in the EViews spreadsheet are a slightly different than those in the UE, p. 502 table,
because the text uses rounded coefficient values for Equation 6.8 and we used non-rounded
EViews estimated coefficients.

YFOLS Y
1995 80.71830 80.30000
1996 82.06108 81.90000
1997 83.65985 83.70000

Step 10. Select Save on the workfile menu bar to save your changes.

Forecasting chicken consumption using a generalized least squares model estimated with the
AR(1) method (UE, 15.2.2):

Complete the section entitled Forecasting chicken consumption using OLS before attempting this
section. The OLS estimate for chicken consumption should already have been estimated and
saved as EQ01 in the Chick6.wf1 workfile and the workfile range & sample should have been
expanded to 1997 and the data for 1995 - 1997 entered into the workfile spreadsheet.

Step 1. Open the EViews workfile named Chick6.wf1.
Step 2. Select Objects/New Object/Equation on the workfile menu bar, enter Y C PC PB YD AR(1)
in the Equation Specification: window, change the Sample: to 1951 - 1994, and click OK.
Step 3. Select Name on the equation window menu bar, enter EQ04 in the Name to identify object:
window, and click OK.
Step 4. Select Forecast on the equation menu bar, enter YFAR1 in the Forecast name: window, set
the Sample range to forecast: to 1951 1997, and click OK.
Step 5. Open YFAR1 & Y in a group window by holding down the Ctrl button, clicking on YFOLS
& Y, selecting Show from the workfile toolbar, and clicking OK. Scroll to the bottom of the
group spreadsheet and make sure that it looks like the table below.
YFAR1 Y
1995 80.05606 80.30000
1996 81.67044 81.90000
1997 83.85621 83.70000
Step 6. Select Save on the workfile menu bar to save your changes.


1
The reason for changing the workfile sample back to the original setting is to ensure that the equation estimation
uses only the data in the original sample (i.e., 1951-1994).
Forecasting chicken consumption using a generalized least squares model estimated with the
Cochrane-Orcutt method(UE, 15.2.2):

The Cochrane-Orcutt method is a multi-step procedure that requires re-estimation until the value
for the estimated first order serial correlation coefficient converges. Follow these steps to use the
Cochrane-Orcutt method to estimate a GLS model for chicken consumption. If you have
questions concerning the procedure, review the appropriate section of Chapter 9. Complete the
section entitled Forecasting chicken consumption using OLS before attempting this section. The
OLS estimate for chicken consumption should already have been estimated and saved as EQ01
in the Chick6.wf1 workfile and the workfile range & sample should have been expanded to 1997
and the data for 1995 - 1997 entered into the workfile spreadsheet.

Step 1. Open the EViews workfile named Chick6.wf1, select Sample on the workfile menu bar,
change the End date from 1997 to 1994, and click OK.
Step 2. Open EQ01 by double clicking its icon in the workfile window. Create a new series for the
residuals (errors) for EQ01, by selecting Procs/Make Residual Series on the EQ01 window
menu bar, enter the name E as the Name for residual series, and click OK.
Step 3. Select Objects/New Object/Equation on the workfile menu bar. Enter E C E(-1) in the
Equation Specification: window and click OK. The coefficient on the E(-1) term (i.e., ), is
positive and significant at the 1% level (t-statistic =3.69 and Prob value =0.0006). This
evidence points to positive serial correlation. Select Name on the equation menu bar, enter EQ02
in the Name to identify object: window, and click OK.
Step 4. Select Objects/New Object/Equation on the workfile menu bar, enter Y-
EQ02.@COEFS(2)*Y(-1) C PC-EQ02.@COEFS(2)*PC(-1) PB-EQ02.@COEFS(2)*PB(-1) YD-
EQ02.@COEFS(2)*YD(-1) in the Equation Specification: window, and click OK. Select Name
on the equation menu bar, enter EQ03 in the Name to identify object: window, and click OK.
2

Step 5. Calculate the new residual series by typing the following formula in the command window:
series E = Y-(EQ03.@COEFS(1) + EQ03.@COEFS(2)*PC + EQ03.@COEFS(3)*PB +
EQ03.@COEFS(4)*YD) and pressing Enter on the keyboard.
Step 6. Re-run EQ02, EQ03 and the series E equation
3
in step 5 sequentially until the estimated (i.e.,
the coefficient on the E(-1) term from EQ02) does not change by more that a pre-selected value
such as 0.001. After 6 iterations, the value for converged (i.e., changed from 0.902484 to
0.902802 between the 5
th
and 6
th
iteration).
Step 7. Convert the constant from the final version of EQ03 by typing the following formula in the
command window: scalar beta0=EQ03.@COEFS(1)/(1-EQ02.@COEFS(2)) and pressing Enter
on the keyboard.
Step 8. Double click the icon in the workfile window and read the value, 26.72580, for the
estimated constant in the lower left of the screen. The final GLS equation for chicken
consumption is Y = 26.72580 - 0.109891*PC + 0.090293*PB + 0.242032*YD, similar to the
truncated coefficient model printed in UE, Equation 9.22, p. 507.

2
Note that the variable names are truncated in the EViews regression output table because they don't fit in the
variable name cell. Nonetheless, the regression is correct. The equation, with the entire variable names printed out,
can be viewed by selecting View/Representations on the equation menu bar.
3
You can re-run an equation by opening the equation in a window, selecting Estimate on the equation menu bar and
clicking OK. You can re-run the series e equation by clicking the cursor anywhere on the equation in the command
window and hitting Enter on the keyboard.
Step 9. Select Sample on the workfile menu bar and change the End date from 1994 to 1997. Open
EQ03 and select Forecast on the equation menu bar. Make sure that Y is checked in the Forecast
of: window. Change the Forecast name: to
YFGLS. Check to make sure that the Sample
range to forecast: is set to 1951 1997, and
click OK (see graphic on the right).
Step 10. Open YFGLS & Y in a group window by
holding down the Ctrl button, clicking on
YFGLS & Y, selecting Show from the
workfile toolbar, and clicking OK. Scroll to
the bottom of the group spreadsheet and
make sure that it looks like the table below.

YFGLS Y
1995 80.05495 80.30000
1996 81.66931 81.90000
1997 83.85514 83.70000

Step 11. Select Save on the workfile menu bar to
save your changes.

Step 12. Note that the predicted values for YFGLS in the EViews spreadsheet are slightly different than
those in the UE, p. 502 table. This is due to the fact that the text uses rounded coefficient values
for Equation 9.22 and we used non-rounded EViews estimated values. Delete this group object
when finished.

Forecasting confidence intervals (UE, 15.2.3):

Follow these steps to make a point prediction of the weight of a male who stands 6'1" tall and
calculate the high and low estimated values for a 95% confidence interval.

Step 1. Open the EViews workfile named Htwt1.wf1.
Step 2. Follow Steps 2 & 3 of the section entitled Forecasting chicken consumption using OLS to
expand the workfile range and sample range from 20 to 21. Double click the X variable icon in
the workfile window, click edit+/- on the series menu bar, scroll to the bottom of the
spreadsheet, replace the NA for the 21
st
observation with 13 (i.e., height in inches above 5'), and
press Enter. To save your changes, click edit+/- on the series menu bar a second time.
Step 3. Select Objects/New Object/Equation on the workfile menu bar, enter Y C X in the Equation
Specification: window, and click OK. Select Name on the equation menu bar and enter EQ01 in
the Name to identify object: window.
Step 4. Select Forecast on the equation menu bar. Enter YF in the Forecast name: window. Check to
make sure that the Sample range to forecast: is set to 1 21, and click OK. To view the forecast
weight of the male student standing 6'1" tall, double click the YF series icon in the workfile
window and scroll to the bottom. The forecast weight for the 21
st
observation is 186.2993.
Step 5. Select Sample on the workfile menu bar, change the Sample range pairs (or sample object to
copy): to 1 20, and click OK.
Step 6. Select Procs/Make Residual Series on the EQ01 window menu bar. Enter the name E as the
Name for residual series, and click OK.
Step 7. Generate a new series for the residuals squared (i.e., E2) by selecting Genr on the workfile
menu bar, entering the equation: E2=E^2 in the Enter equation: window, and clicking OK.
Step 8. Generate a new series named XDEV2 for the residuals squared by selecting Genr on the
workfile menu bar, entering the equation: XDEV2=(x-@mean(x))^2 in the Enter equation:
window, and clicking OK.
Step 9. To calculate the upper confidence interval for the 6'1" student, type the following formula (all
one equation) in the command window, and press Enter on the keyboard:
scalar YF_HIGH=Yf(21)+((@sum(E2)/@obs(E2)*(1+(1/@obs(E))+((X(21)-
@mean(X))^2/@sum(XDEV2))))^.5)*@qtdist(.975,(eq01.@regobs-eq01.@ncoef))

Double click YF_HIGH in the workfile window to view the upper confidence interval in the
lower left of the screen (i.e., ).

Step 10. To calculate the lower confidence interval for the 6'1"student, type the following formula
(all one equation) in the command window and press Enter on the keyboard:
scalar YF_LOW=Yf(21)-((@sum(E2)/@obs(E2)*(1+(1/@obs(E))+((X(21)-
@mean(X))^2/@sum(XDEV2))))^.5)*@qtdist(.975,(eq01.@regobs-eq01.@ncoef))

Double click YF_LOW in the workfile window to view the upper confidence interval in the lower
left of the screen (i.e., )

Forecasting with simultaneous equation systems (UE 15.2.4):

EViews has a Type of Object: named Model that enables you to solve a system of simultaneous
equations and use the model for forecasting and simulation. EViews models do not contain
unknown coefficients to be estimated. Instead, the Model object allows you to solve for
unknown values for the endogenous variables. For a description of various methods of creating
model objects in EViews and techniques for using these objects to forecast and perform
simulations see Help/Contents/Hypothesis Testing and Forecasting/Model Solve (Forecast
and Simulation).

This topic is beyond the scope of this guide but, if you were to do it, you would follow these
steps to forecast with a simultaneous equation model:

Step 1. Create a model by selecting Object/New Object/Model in the main toolbar, type in a name
for your model in the Name for Object: window, click OK, and enter the previously estimated
equations in the model window.
Step 2. To solve the model, simply select the Solve button in the model toolbar. You should see the
Model Solution dialog box offering various options for controlling the solution process. EViews
solves for the endogenous variables, given data for the exogenous variables.
Forecasting with ARIMA models (UE 15.3):

ARIMA (autoregressive, integrated, moving average) models use three tools for modeling the
serial correlation in the disturbance:
1. The first tool is the autoregressive or AR term. The AR(1) model introduced above uses only
the first-order term but, in general, you may use additional, higher-order AR terms. Each AR
term corresponds to the use of a lagged value of the residual in the forecasting equation for
the unconditional residual. An autoregressive model of order p is denoted as AR(p).
2. The second tool is the integration order term. Each integration order corresponds to
differencing the series being forecast. A first-order integrated component means that the
forecasting model is designed for the first difference of the original series. A second-order
component corresponds to using second differences, and so on.
3. The third tool is the MA, or moving average term. A moving average forecasting model uses
lagged values of the forecast error to improve the current forecast. A first-order moving
average term uses the most recent forecast error, a second-order term uses the forecast error
from the two most recent periods, and so on.

In ARIMA forecasting, you assemble a complete forecasting model by using combinations of the
three building blocks described above. You can use the correlogram view of a series for this
purpose (see Testing for nonstationarity by calculating the auto correlation function ACF for a
description of this process).

UE did not present an example of ARIMA estimation, but one will be estimated here to
demonstrate how to estimate an ARIMA model in EViews. In Chapter 12 we determined that CO
was integrated order one and probably followed an AR(1) process. For purposes of illustration,
follow these steps to estimate an ARIMA(1,1,2)
4
model for CO:

Step 1. Open the EViews workfile named Macro14.wf1.
Step 2. Select Objects/New Object/Equation on the workfile menu bar, enter D(CO) C AR(1)
MA(1) MA(2) in the Equation Specification: window, and click OK to view the EViews
Estimation Output printed below.

Step 3. Select Name on the
equation menu bar and
enter EQ01 in the Name to
identify object: window.
Step 4. Follow Steps 2 & 3 of
the section entitled
Forecasting chicken
consumption using OLS to
expand the workfile range
and sample range from
1994 to 1998.
Step 5. Select Forecast on the
EQ01 menu bar. Enter COF
in the Forecast name:
window. Check to make
sure that the Sample range
to forecast: is set to 1998,
and click OK.
Step 6. To view the forecast, double click the COF series and scroll to the bottom of the spreadsheet
to view the following forecast values for 1995 - 1998:

COF
1995 3536.473
1996 3604.213
1997 3671.953
1998 3739.694


4
ARIMA(1,1,2) represents a model that has a first order autoregressive term (i.e., AR(1)), is integrated order one
and has two moving average terms (i.e., MA(1) & MA(2)).
Dependent Variable: D(CO)
Method: Least Squares
Date: 07/15/00 Time: 10:35
Sample(adjusted): 1965 1994
Included observations: 30 after adjusting endpoints
Convergence achieved after 100 iterations
Backcast: OFF (Roots of MA process too large for backcast)
Variable Coefficient Std. Error t-Statistic Prob.
C 67.74008 3.655560 18.53070 0.0000
AR(1) 0.718868 0.251279 2.860836 0.0082
MA(1) -0.943222 0.362834 -2.599598 0.0152
MA(2) -0.773132 0.292764 -2.640801 0.0138
R-squared 0.588152 Mean dependent var 72.04333
Adjusted R-squared 0.540631 S.D. dependent var 39.20346
S.E. of regression 26.57083 Akaike info criterion 9.521071
Sum squared resid 18356.23 Schwarz criterion 9.707897
Log likelihood -138.8161 F-statistic 12.37671
Durbin-Watson stat 2.201209 Prob(F-statistic) 0.000032
Inverted AR Roots .72
Inverted MA Roots 1.47 -.53
Estimated MA process is noninvertible
Exercises:

2.
a.
b. Refer to Running a simple regression for Woody's Restaurants example to estimate
Equation 3.7 and Forecasting chicken consumption using OLS to make an unconditional
forecast of the gross sales volume at the three proposed restaurant sites.
c. Open the EViews workfile named Fish8.wf1, select Objects/New Object/Equation on
the workfile menu bar, enter F C PF/PB log(YD) P in the Equation Specification:
window, and click OK. Then follow the steps in Forecasting chicken consumption using
OLS to make an unconditional forecast of F for 1971-1974, given the numbers printed in
the table in UE, Exercise 2c, p.517.

3. Open the EViews workfile named Htwt1.wf1, select Objects/New Object/Equation on the
workfile menu bar, enter Y C X in the Equation Specification: window, and click OK. Ask
three students their height (X=the height in inches exceeding 5'). Then follow the steps in
Forecasting chicken consumption using OLS to make the Unconditional forecast of the
weight (YF) for each student in your sample. Ask these students their actual weight (Y) and
compare Y with YF.

4.
a. Follow the steps in Forecasting confidence intervals to calculate the 95% confidence
interval for the weight for a person standing 5'9" tall (i.e., X = 9).
b.

6. Follow the steps in Chapter 1 for each part of this exercise.